Robust and optimal laser cooling of trapped ions
J. CcrrilloQ A. Rctzkcr, and M. B. Plcnio
Institute for Mathematical Sciences, Imperial College London, SW7 2PG, UK and
QOLS, The Blackett Laboratory, Imperial College London, Prince Consort Rd., SW7 2BW, UK
We present a robust and fast laser cooling scheme suitable for trapped atoms and ions. Based
on quantum interference, generated by a special laser configuration, it is able to rapidly cool the
system such that the final phonon occupation vanishes to zeroth order in the Lamb-Dicke parameter
in contrast to existing cooling schemes. Furthermore, it is robust under conditions of fluctuating
laser intensity and frequency, thus making it a viable candidate for experimental applications.
OS
o
o
(N
^— > ■
c ■
a ■
=5
cr
>
SO
00
in
o
Os
O
X
Introduction — Laser cooling is a crucial ingredient
for probing quantum properties of matter
is a key factor in a wide variety of experiments ranging
from Bosc- Einstein condensates and quantum computing
to quantum simulation with atoms and ions. Variants
of coolingschemes range from Doppler cooling for free
particles m and its partner, side band cooling for bound
particles 5|, @| , to dark state cooling schemes for free Q
and bound particles 0] relying on quantum interference
which arises thanks to their non-trivial internal electronic
structures.
At present, sideband cooling is the method of choice
for trapped ions. It is a necessary requirement for effi-
cient cooling that motional sidebands with frequency v
can be resolved, i.e., the linewidth of the optical tran-
sition 7 <C v. Cooling is then achieved by the red side-
band transition which excites the atom while at the same
time annihilating a phonon to ensure energy conserva-
tion. This transition rate must be higher than that on the
carrier and blue sideband transitions that heat the sys-
tem either through recoil after spontaneous decay (car-
rier) or coherent generation of a phonon (blue sideband).
The selection of the red transition is ensured by the ro-
tating wave approximation, i.e. energy conservation, as
long as the Rabi frequency fl of the laser satisfies £1 <C v.
One method to suppress the carrier and blue sideband
transitions employs destructive interference exhibited for
example in dark states For this reason, Electromag-
netically Induced Transparency (EIT) [§] has become an
inspiration for a variety of proposed laser cooling schemes
for trap ped ions such as EIT cooling [n| and Stark-shift
cooling llj. In EIT cooling, interference eliminates the
carrier transition to improve cooling performance while
in the Stark-shift scheme this is achieved in a rotated
sorption spectrum, while constraining the blue sideband
to a region with negligible absorption, thus achieving low
final temperatures. The final state of the system is then
basis in a suitable interaction picture 12 1.
In EIT cooling the existence of a dark state of
a three level system allows final temperatures below
fvy/kB- This is achieved in a three level scheme subject
to Raman lasers with strong blue single-photon detuning
that couple both the ground state and a meta-stable state
to an excited dissipative state. Among the dressed states
of the system there is one dark state that cancels the car-
rier transition. Well chosen parameters, can center the
red sideband transition under a peak of the Fano-like ab-
p (B/T) = \dark)(dark\ ® V]a n |n)(n| + o(r/ 2
(1)
where \dark) is the internal level steady state and \n)
are the number states of the external degrees of free-
dom (dof), with a final mean phonon number of order
( 7 /4|A|) 2 0.
In Stark-shift cooling [ll[ on the other hand one laser
drives transitions between the ground and a meta-stable
state and another two resonant Raman lasers couple a
superposition of both to the excited state. This first
laser generates Rabi oscillations between the dark state
and the orthogonal bright state. If the coupling is cor-
rectly tuned, the oscillations will also involve neigh-
boring mechanical levels so that states \dark) \n) and
\bright) \n — 1) are coupled while carrier transitions are
eliminated. As the EIT scheme, Stark-shift cooling
achieves a final temperature that is in leading order in-
dependent of the Lamb-Dicke parameter 77.
It is now natural to ask whether these schemes can
be combined to reach vanishing temperature in lead-
ing order in the Lamb-Dicke parameter. As this work
demonstrates, this is indeed possible. We present here
a novel cooling scheme, a judicious combination of EIT
and Stark-shift cooling, which share the same dark state.
This in turn allows for both the carrier and the blue side-
band transitions, and hence heating, to be suppressed by
interference. As a result the final temperature, to zeroth
order in the Lamb-Dicke parameter, vanishes. To the
best of our knowledge this is the first scheme that is able
to achieve this for trapped particles.
That the combination of EIT and Stark-shift cooling
is successful, may seem surprising. A classical analy-
sis would lead one to expect that the efficiency of the
combined scheme to be the arithmetical average of the
individual ones. It is the quantum nature of the inter-
action however that makes them interfere constructively
in such a way as to outperform its constituent schemes.
The quantum interference can be designed to completely
cancel the two leading heating processes in the system
2
FIG. 1: A three level structure made up of \gi) and \g2) and
an excited dissipative state |e) which decays at rate 7. The
ground levels are coupled to the excited state by a pair of
Raman beams under a detuning A whose Rabi frequencies
are Q.a and mechanically interact with the ion with opposite
Lamb-Dicke parameters ija- The ground levels are directly
coupled by an effective Rabi frequency Qb and a Lamb-Dicke
parameter t]b.
resulting in the state
p (robust) = | dflrfc ^ ^ arfc | ^ | ) (0| + ( 7 f) (2)
with a final temperature that vanishes in leading order
in the Lamb-Dicke parameter. Remarkably we will also
demonstrate that the scheme also exhibits a significant
improvement with regards to the stability under fluctuat-
ing laser intensities. This implies that it remains robust
under general experimental conditions, thus assuring its
feasibility.
Description of robust cooling - The combined scheme,
as presented in fig. [TJ involves a trapped ion with an in-
ternal electronic structure made up of a ground state and
a metastable state \gi) and | (72) and an excited dissipa-
tive state |e). For simplicity, we will refer to the coupling
between the lower states \gi) and | (72) as "SSh coupling"
and between the g-manifold and the excited state |e) as
"EIT couplings" . A harmonic well models the trap po-
tential. It is characterized by its equally spaced levels |n)
representing the Fock state of n phonons and the creation
and annihilation operators b and b^ . Stark-shift and EIT
coolings can be regarded as particular instances of this
scheme when t]a — > and — > respectively.
The quantum theory of laser cooling of trapped parti-
cles was developed in [l3|, 14, 15| and we will follow the
notation in |15| . Expanding the Hamiltonian up to first
order in the Lamb-Dicke parameters, it can be split into
three contributions, the trap potential and the internal
dof:
H = vb^b - 2J Si \gi) (g l \ ,
(3)
where are S^s are the detunings with respect to the lasers.
The interaction of the laser beams with the ion due to
the EIT part:
H
EIT
n A «< e + << e ) + «< e - of •«) (b + &t) ( 4 )
and the interaction of the laser beams with the ion due
to the Stark-shift part:
Hssh = finof > g *v A Vt A + '//;<?/,<;•"<••■:/, + (5)
where a
(m,n)
= (\m)(n\ + h.c.) and a y m,n ^ =
(i\m) (n\ + h.c).
The physics of the scheme can best be understood by
extending the analysis of the steady state Eq.© to the
next order in the Lamb-Dicke parameter. By inspection
of the Hamiltonian one can consider the following non-
normalized pure state:
l*> = - <72>|0) -it]B\gi +52)|1) + o(r l 2 ) (6)
Under the effect of the EIT couplings (Eq. [4j both terms
of the superposition interfere destructively and as a re-
sult | I F) remains invariant. However |\&) is not invariant
under the free Hamiltonian (Eq. ([3])) which introduces a
phase shift between the two components. For a suitable
tuned parameters, the Stark shift coupling part (Eq. [5])
can cancel the effect of the free Hamiltonian and as a
consequence \^f) will be a dark state of the Liouvillian as
it does not suffer any spontaneous emission losses cither.
This is achieved when
o,
(7)
From an experimental point of view, it is worth noting
that this resonance condition is characterized by the quo-
tient of the Lamb-Dicke parameters. These can be set up
at the beginning of the experiment to a high precision.
Losses from the excited level are incorporated in the
master equation:
at
H EIT + H SShlP }+C d p
(8)
where C d contains the spontaneous emission of the ex-
cited level:
C d =
E
le,i^i,ePe,i<^e,i ~ PO~e,e ~ G e ,eP (9)
where Uj^k = \j) (k\ and ~p ~\ dxe lkx pe~ lkx .
After expansion of the rest of the terms up to second
order in the Lamb-Dicke parameter and adiabatic elimi-
nation of the internal dof we find the rate equation:
7 n,n
a Pext 17 1 iw a ri+l.n+1 a n,n\
= \\n + l){A-p ext - A +Pe < xt )
dt
/ a n — l,n—l
n \ A +Pext
A -Pext)\
In the spirit of [15( the rates A± can be expressed
A± = 2Re[D + S(tv)}
(10)
111)
where D is the diffusion coefficient due to spontaneous
emission from the excited atomic states. Here D = as
3
3.5
2.5
0.5 -
'7.4
i-©.io
-0.2 -0.15 -0.1 -0.05 0.05 0.1 0.15 0.2
FIG. 3: Expected value of occupation (n) as a function of the
Lamb-Dicke parameter quotient and of variations around the
optimal Rabi frequency £Ib , where it is shown how variations
have a worse effect as r\A vanishes, that is, as the Stark-shift
cooling regime is approached.
FIG. 2: The deviation of the analytical from the exact nu-
merical results in the final population is plotted versus the
deviation in Qb (in units of trap frequency) from the optimal
operating point for different Lamb-Dicke parameters.
the population of the excited states vanishes due to the
dark state nature of the final state. S(v) is the fluctua-
tion spectrum of Heisenberg operator F(t):
S( V )
1
2Mv
dte wt (F(t)F(0)}
(12)
a y ie ~ a y 2e and
where F = Feit + Fssh and Feit
Fssh = TiB^B^y 192 are the part in the interaction Hamil-
tonian Eq. 0] and Eq. [5] that multiply 6 + fe T . The aver-
age can be calculated using the quantum regression the-
orem. We may decompose the operator F into compo-
nents of EIT and Stark-shift so that the overall heat-
ing rate can then be split into three parts: the EIT
part A+ IT , the Stark-shift part A+ Sh and interaction be-
tween EIT and the Stark-shift part AT* for the remain-
(r, A (v + 2{l B )) 2 /V,
ing cases. Their results are: A+ IT
j^SSh
- (f]B^Br/V and A
where ^
-2r lA r lB (v+2VL B )Sl B /V
2fiir
T> r 2 (i/+2J2 B ) 2 + (-2n^ + (iy+20 B )(A+iy+0 B )) 2
A+ = \i lA { v + 2n B )-i lB n B ] 2 /V
Then
(13)
which will be identically zero for the condition in Eq.([7]).
This emphasizes the role of the interaction between the
EIT and Stark-shift cooling in order to assure the ground
state is populated with probability 1 . The mean occupa-
tion number n is given by
- , thus yielding
it
0.
A--A+ ■
Fig. [2] shows the numerical results of the approach to the
(n) = point as the Lamb Dicke parameter goes to zero.
Robustness — The constructive interference between
EIT and Stark-shift contribution is also crucial for under-
standing the robustness of the scheme under fluctuating
parameters. If the Rabi-frcqucncies deviate from Eq.0
by, Afl A / B , the final populations as
(n) oc {An A ) 4 (An B ) 2
(14)
instead of second order as is usually the case. As is exem-
plified in fig. [3J under a given value of the fluctuations of
the laser intensities, the final mean occupation decreases
abruptly as one moves away of the Stark-shift only or EIT
only regimes. This guarantees promising performance
under real experimental conditions, overcoming the main
drawback of the previous dark-state cooling schemes.
Rate — The interference structure of this scheme is
also essential for the high cooling rate W = A_ — A + .
For the resonance condition eq. ([7]) we find
W
8Tr,j
+ (i/-2n B )(A
n B )y + r 2 (v
2tt B y
(15)
£l B is the only Rabi frequency involved in the condition
Eq.©. As a function of fl B , Eq. (fT5"l) takes the approx-
imate shape of a squared Lorentzian, with a peak close
to fl B = v/2 at which point the cooling rate expression
reduces to
W
(16)
This is also an optimal point for Stark-shift cooling [ll|
but it should be noted that the cooling rate of Robust
cooling is slightly higher than that of Stark-shift cool-
ing. Since the internal dynamics should be much faster
than the external one for the perturbation approach to
work the analytic result is not valid for VL A < r/ A v, which
means that the cooling can be as fast as 1 order of mag-
nitude less than the trap frequency. Since the final state
(eq. [6]) is a pure state up to second order in the Lamb-
Dicke parameter there is a simple unitary rotation such
that the final state has vanishing number of phonons to
fourth order in the Lamb Dicke parameter.
Implementation — The way the effective fl B and r] B
couplings can be physically implemented is not unique.
4
One way is to use two lasers to create the EIT cooling
part and to use magnetic gradients for the Stark-
shift part. In this system the magnetic gradients create
a coupling of the following type: Xa z (b + b'), where A is
proportional to the magnetic gradients and the TLS is
driven using a microwave: £ldO~ x cos Wdt, where ild corre-
sponds to the Rabi frequency of the driving and Ud to the
driving frequency. After a polaron transformation the re-
sulting Hamiltonian is exactly as in Eq. [5] when the Rabi
frequency is replaced by fid and the Lamb Dicke param-
eter is replaced by K
This scheme can be especially useful to cool nano scale
resonators, by using the setup described in [17|. In this
setup an NV center is coupled to a diamond cantilever,
the coupling is achieved by magnetic gradients resulting
in the same Hamiltonian as described above. In can-
tilevers the speed of cooling is very important due to the
finite Q value, which is an important factor limiting the
achievable final temperatures at present. The high cool-
ing rate achieved by the described scheme will result in
lower final temperatures bringing us closer to the goal of
achieving the quantum regime in cantilever systems.
Alternatively, one can also use Raman beams with
large single-photon detuning to couple levels \gi) and
| (72 ) - By adiabatically eliminating the upper level the re-
lationships between our effective parameters fls and tjb
and the physical values Q p and r\ v are found to be £Ib =
ft 2
p
A
2A^
and flBTjB = ^pVp A{A^-l'i)
Neglecting the rf 1 correction in the first expression we find
Vb = Vphys ^"r^ wmcn yields r] B = 2n p for sufficiently
large detunings.
If we chose the optimal point for the cooling rate and
fluctuations Qb = f , the condition becomes r} p jr\A = 4.
This can be achieved for a layout where beam B is col-
inear to the trap axis and the beam A is 60° away from
the axis. Even if the optimal situation is for an angle
separation of 60°, the robustness of the scheme ensures
excellent performance for any geometrical configuration.
Proposed experimental layouts [2(| use vacuum cham-
bers with windows at 22.5° and/or 45° from the trap
axis, which generally allow for an angle range of about
±10°. Taking 45° as an operating value, the Lamb-Dickc
parameter quotient becomes kb/va = 2y/2. Different
optimal set of parameters can be obtained for this situ-
ation depending on whether the cooling rate or the final
temperature want to be optimized. For the latter, con-
dition Eq.0 has to be observed, ft a = 0.6^ and A ~ 0.
This will assure an extremely stable cooling rate which is
generally 2 orders of magnitude smaller than EIT cooling
but still better than sideband cooling. This final result
can be improved depending on the particular values of
the transition lincwidth T. On the contrary, if the cool-
ing rate is to be enhanced, condition Eq.((7|) won't be
satisfied. In particular, for £Ia — 0.4z/, £1b ~ 0.45^ with
5% fluctuation and A ~ — 2v the population can still be
as low as 10 _3 while having a cooling rate several orders
of magnitude over that of EIT cooling. Taking into ac-
count the fact that angles up to 55° are accessible the
cooling rate can still be improved by up to two orders of
magnitude.
Multi mode cooling — Finally, the cooling scheme has
also been tested for an ion chain using Monte Carlo sim-
ulation
It
191 ]. The robustness of the scheme implies a
wide range of operational Rabi frequencies or, in a dif-
ferent perspective, a wide range of trap frequencies for
a given Rabi frequency. Thus, a particular central mode
frequency can be addressed so that also the neighboring
modes benefit from the cooling. In a multi-mode environ-
ment with up to 3 ions promising results have been ob-
tained and a detailed study will be presented elsewhere.
Conclusion — To conclude, we have introduced a cool-
ing scheme that cools to zero temperature without any
corrections in zeroth order in the Lamb Dicke parameter.
Beyond the academic interest of proving the existence of
such a scheme, its robustness makes it extremely attrac-
tive for experimental realizations.
Acknowledgment — J. C. acknowledges support from
the AXA Research Fund. A.R. acknowledges the support
of EPSRC project number EP/E045049/1 and MBP ac-
knowledges support from the Royal Society and the EU
STREP project HIP.
[9
[10
[11
[12
[13
[14
[15
Electronic address: j.cerrillo@imperial.ac.uk
S. Chu, Rev. Mod. Phys. 70, 685 (1998).
C. N. Cohen- Tannoudji, Rev. Mod. Phys. 70, 707 (1998).
W. D. Phillips, Rev. Mod. Phys. 70, 721 (1998).
T. Hansch and A. Schawlow, Opt. Commun 13, 68
(1975).
D. Wineland and H. Dehmelt, Bull. Am. Phys. Soc 20,
637 (1975).
D. J. Wineland, R. E. Drullinger, and F. L. Walls, Phys.
Rev. Lett. 40, 1639 (1978).
A. Aspect, E. Arimondo, R. Kaiser, N. Vansteenkiste,
and C. Cohen- Tannoudji, Phys. Rev. Lett. 61, 826
(1988).
R. Dum, P. Marte, T. Pellizzari, and P. Zoller, Phys.
Rev. Lett. 73, 2829 (1994).
M. Fleischhauer, A. Imamoglu, and J. P. Marangos, Rev.
Mod. Phys. 77, 633 (2005).
G. Morigi, J. Eschner, and C. H. Keitel, Phys. Rev. Lett.
85, 4458 (2000).
A. Retzker and M. B. Plenio, New J. of Phys. 9, 279
(2007).
D. Jonathan, M. B. Plenio, and P. L. Knight, Phys. Rev.
A 62, 042307 (2000).
M. Lindberg and J. Javanainen, J. Opt. Soc. B 3, 1008
(1986).
J. Javanainen and S. Stenholm, Applied Physics A 24,
151 (1981).
J. I. Cirac, R. Blatt, and P. Zoller, Phys. Rev. A 46, 2668
(1992).
5
[16] F. Mintert and C. Wunderlich, Phys. Rev. Lett. 87,
257904 (2001).
[17] P. Rabl, P. Cappellaro, M. V. G. Dutt, L. Jiang, J. R.
Maze, and M. D. Lukin, Phys. Rev. B 79, 041302 (2009).
[18] K. Molmer, Y. Castin, and J. Dalibard, J. Opt. Soc. Am.
B 10, 524 (1993).
[19] M. B. Plenio and P. L. Knight, Rev. Mod. Phys. 70, 101
(1998).
[20] personal communication with K. Singer and T. Schatz.